Kernel-based Algorithm Research Articles

Optimizing cervical cancer classification using transfer learning with deep gaussian processes and support vector machines

BackgroundCervical cancer is the fourth most frequent cancer in women worldwide. Even though cervical cancer deaths have decreased significantly in Western countries, low and middle-income countries account for nearly 90% of cervical cancer deaths. While Western countries are leveraging the powers of artificial intelligence (AI) in the health sector, most countries in sub-Saharan Africa are still lagging. In Uganda, cytologists manually analyze Pap smear images for the detection of cervical cancer, a process that is highly subjective, slow, and tedious. Machine learning (ML) algorithms have been used in the automated classification of cervical cancer. However, most of the MLs have overfitting limitations which limits their deployment, especially in the health sector where accurate predictions are needed.MethodsIn this study, we propose two kernel-based algorithms for automated detection of cervical cancer. These algorithms are (1) an optimized support vector machine (SVM), and (2) a deep Gaussian Process (DGP) model. The SVM model proposed uses an optimized radial basis kernel while the DGP model uses a hybrid kernel of periodic and local periodic kernel.ResultsExperimental results revealed accuracy of 100% and 99.48% for an optimized SVM model and DGP model respectively. Results on precision, recall, and F1 score were also reported.ConclusionsThe proposed models performed well on cervical cancer detection and classification, and therefore suitable for deployment. We plan to deploy our proposed models in a mobile application-based tool. The limitation of the study was the lack of access to high-performance computational resources.

A Kernel-Based Calibration Algorithm for Chromatic Confocal Line Sensors.

In chromatic confocal line sensors, calibration is usually divided into peak extraction and wavelength calibration. In previous research, the focus was mainly on peak extraction. In this paper, a kernel-based algorithm is proposed to deal with wavelength calibration, which corresponds to the mapping relationship between peaks (i.e., the wavelengths) in image space and profiles in physical space. The primary component of the mapping function is depicted using polynomial basis functions, which are distinguished along various dispersion axes. Considering the unknown distortions resulting from field curvature, sensor fabrication and assembly, and even the inherent complexity of dispersion, a typical kernel trick-based nonparametric function element is introduced here, predicated on the notion that similar processes conducted on the same sensor yield comparable distortions.To ascertain the performance with and without the kernel trick, we carried out wavelength calibration and groove fitting on a standard groove sample processed via glass grinding and with a reference depth of 66.14 μm. The experimental results show that depths calculated by the kernel-based calibration algorithm have higher accuracy and lower uncertainty than those ascertained using the conventional polynomial algorithm. As such, this indicates that the proposed algorithm provides effective improvements.

Method for noninvasive HV/MV switchgear motion analysis using kernel-based algorithm with adaptive feature extraction

When designing, developing, and testing medium-voltage (MV) and high-voltage (HV) switchgears, it is of utmost importance to analyze the movements of their mechanical parts, such as drive trains, contact nozzles, etc. This ensures the safety of switchgear operations and enables detecting and predicting issues that could lead to component damage, power interruptions, reduced efficiency, or even switch failure. Conventional invasive measuring setups, including encoders and laser distance measurements, are often difficult or expensive to use, due to undesirable environmental properties such as high temperatures and/or high voltages, or dimensional constraints often encountered in testing laboratories, compact substations or confined equipment rooms. Video object tracking can be a viable solution in such conditions, allowing for the extraction of the trajectory of mechanical parts of an object under analysis. This paper proposes a novel kernel-based algorithm with adaptive feature extraction for precise colour-based object tracking through video processing. The implemented method is based on the principles of the CamShift algorithm, augmented with fusion with the Kalman filter for continuous estimation and prediction of the object’s position based on the available measurements while reducing sensitivity to noise. Beyond precise tracking of the object of interest, the algorithm automatically adapts the mask used in the standard CamShift algorithm by extracting and processing the features of the selected object. This approach exhibits flexibility and robustness in adverse industrial environments. It supports modifications according to user preferences and represents a cost-effective alternative to conventional methods, while performing real-time processing. Experimental results underscore this noninvasive approach is flexible, highly robust, and enables tracking of coloured markers even in challenging conditions like poor lighting, significant blur, and low frame rates.

Machine Learning of Reactive Potentials.

In the past two decades, machine learning potentials (MLPs) have driven significant developments in chemical, biological, and material sciences. The construction and training of MLPs enable fast and accurate simulations and analysis of thermodynamic and kinetic properties. This review focuses on the application of MLPs to reaction systems with consideration of bond breaking and formation. We review the development of MLP models, primarily with neural network and kernel-based algorithms, and recent applications of reactive MLPs (RMLPs) to systems at different scales. We show how RMLPs are constructed, how they speed up the calculation of reactive dynamics, and how they facilitate the study of reaction trajectories, reaction rates, free energy calculations, and many other calculations. Different data sampling strategies applied in building RMLPs are also discussed with a focus on how to collect structures for rare events and how to further improve their performance with active learning.

UBES: Unified scatter correction using ultrafast Boltzmann equation solver for conebeam CT

A semi-analytical solution to the unified Boltzmann equation is constructed to exactly describe the scatter distribution on a flat-panel detector for high-quality conebeam CT (CBCT) imaging. The solver consists of three parts, including the phase space distribution estimator, the effective source constructor and the detector signal extractor. Instead of the tedious Monte Carlo solution, the derived Boltzmann equation solver achieves ultrafast computational capability for scatter signal estimation by combining direct analytical derivation and time-efficient one-dimensional numerical integration over the trajectory along each momentum of the photon phase space distribution. The execution of scatter estimation using the proposed ultrafast Boltzmann equation solver (UBES) for a single projection is finalized in around 0.4 seconds. We compare the performance of the proposed method with the state-of-the-art schemes, including a time-expensive Monte Carlo (MC) method and a conventional kernel-based algorithm using the same dataset, which is acquired from the CBCT scans of a head phantom and an abdominal patient. The evaluation results demonstrate that the proposed UBES method achieves comparable correction accuracy compared with the MC method, while exhibits significant improvements in image quality over learning and kernel-based methods. With the advantages of MC equivalent quality and superfast computational efficiency, the UBES method has the potential to become a standard solution to scatter correction in high-quality CBCT reconstruction.

Simple yet efficient downscaling of land surface temperatures by suitably integrating kernel- and fusion-based methods

Kernel-based and fusion-based methods have been widely used to downscale satellite-derived land surface temperatures (LSTs) for obtaining LSTs with high spatiotemporal resolution. The integration of these two methods can potentially provide a more effective method for LST downscaling. However, a downscaling algorithm that appropriately integrates the advantages of these two methods in accuracy and stability, yet is characterized by adequate simplicity, effectiveness, and global applicability, is still lacking. To address this issue, here we proposed the Simple and Effective Downscaling (SED) algorithm, which integrates kernel- and fusion-based methods suitably and uses a single adjacent Landsat image to downscale MODIS LSTs by statistical regression. The SED was tested in 50 different study areas worldwide, and its performance was compared with four benchmark downscaling algorithms, including a kernel-based algorithm (the RFT21), a fusion-based algorithm (the FSDAF), and the simple average (the Hybrid 1) and weighted average (the Hybrid 2) of the RFT21 and the FSDAF. Our evaluations show that the SED performs better in almost all study areas and has finer global applicability when compared with these four benchmark algorithms. The mean RMSE of the SED is 1.29 K over these 50 study regions, with an accuracy improvement of over 13 % relative to the best benchmark algorithm. Particularly, the SED exhibits higher accuracies in tropical and temperate regions, over urban areas and savannas, and when using Landsat images closer to MODIS LSTs. We believe that the SED can contribute to the generation of high-quality and high spatiotemporal resolution LSTs over global lands.

Pixel-level image analysis to derive the broncho-artery (BA) ratio employing HRCT scans: A computer-aided approach

Bronchiectasis in children is a major health issue which can be life-threatening if not diagnosed and effectively treated. In the diagnosis of bronchiectasis, an increased broncho-arterial (BA) ratio is considered a significant marker. The BA ratio is measured by evaluating BA pairs, using high-resolution computed tomography (HRCT) scans. Detecting BA pairs automatically is challenging due to the complex characteristics of BA pairs and the ambiguous appearance of the bronchi. This study proposes an effective computerized approach to detect BA pairs and assess BA ratio using HRCT scans of children and employing computer-aided techniques and novel custom-build algorithms. Attention is given to reconstructing broken bronchial walls and identifying discrete BA pairs using custom-built kernel based and patch-based algorithms for pixel-level image analysis. To detect BA pairs, the lung region is segmented in the HRCT slices and image preprocessing techniques, including noise reduction, binarizing, largest contour detection and a hole-filling algorithm, are applied. A histogram analysis method is introduced to clean the images. A kernel-based algorithm is proposed to reconstruct the pixel distribution if the bronchial wall is so that the bronchi can be detected precisely. Potential arteries are detected using balanced histogram thresholding, morphological opening and an approach based on four conditions related to the object area circularity, rectangular boundary box ratio and enclosing circle area ratio. Potential bronchi are detected through matching of object coordinates with potential arteries, hole-filling and four condition based approaches. The potential BA pairs are detected by matching the coordinates of potential bronchi with those of potential arteries as the artery and bronchus are adjacent to each other in BA pairs. Finally, from the potential BA pairs, actual BA pairs are identified using a custom-built patch algorithm. The study is conducted using 2471 HRCT slices of seven children, obtained from the Royal Darwin Hospital, Australia. The BA ratio is derived based on the ratio of diameters, major axis lengths, minor axis lengths, area, convex hull and equivalent diameter where the BA ratios are respectively 0.51–0.65, 0.49–0.59, 0.59–0.77, 0.25–0.42, 0.29–0.47, 1.5–2 and 0.50–0.65.

Root cause diagnosis of plant-wide oscillations based on fuzzy kernel multivariate Granger causality

BackgroundPlant-wide oscillations are commonly observed in industrial processes and can have significant impacts on product quality and energy consumption. Accurately diagnosing the root causes of these oscillations is essential to mitigate their effects. However, when these oscillations occur, the resulting data often exhibits strong nonlinearity, making root cause diagnosis a challenging task. MethodIn this study, we propose the use of kernel multivariate Granger causality (KMGC) test as an analytical tool for root cause diagnosis. The kernel function allows for low-dimensional data to be mapped to a higher-dimensional space, making it easier to process data with strong nonlinearity. To address the computational complexity associated with complex kernel functions, we have combined the KMGC with a fuzzification function to create the fuzzy kernel multivariate Granger causality (FKMGC) analysis method. This approach improves computational efficiency, particularly in industrial scenarios where Gaussian kernel functions are commonly used. Significant findingsOur simulation and plant-wide oscillation case study demonstrate that FKMGC accurately identifies the root cause of oscillations. Furthermore, FKMGC significantly reduces analysis time compared to Gaussian kernel-based algorithms, with a 95% reduction in running time observed in our experiments. These findings suggest that FKMGC is an effective and efficient tool for diagnosing the root causes of plant-wide oscillations in industrial settings.

Kernel-based clustering via Isolation Distributional Kernel

Clustering has become one of the widely used automatic data-labeling techniques applied in a variety of disciplines. Kernel-based clustering is a technique designed to identify non-linearly separable clusters with irregular shapes. However, existing kernel-based clustering algorithms usually produce weaker clustering outcomes than density-based clustering, and they have high computational cost which limits their applications to small datasets only. In this paper, we contend that these limitations are mainly due to the use of (a) the Expectation-and-Maximization algorithm as an optimization procedure, and (b) a non-adaptive kernel. In addition, to address the limitations of current kernel-based algorithms, we propose the first clustering algorithm that employs an adaptive distributional kernel without any optimization, while achieving a similar optimization objective function. We demonstrate its superior performance of identifying complex clusters on massive datasets under different real-world application scenarios.

Semi-Supervised Heterogeneous Domain Adaptation: Theory and Algorithms.

Semi-supervised heterogeneous domain adaptation (SsHeDA) aims to train a classifier for the target domain, in which only unlabeled and a small number of labeled data are available. This is done by leveraging knowledge acquired from a heterogeneous source domain. From algorithmic perspectives, several methods have been proposed to solve the SsHeDA problem; yet there is still no theoretical foundation to explain the nature of the SsHeDA problem or to guide new and better solutions. Motivated by compatibility condition in semi-supervised probably approximately correct (PAC) theory, we explain the SsHeDA problem by proving its generalization error - that is, why labeled heterogeneous source data and unlabeled target data help to reduce the target risk. Guided by our theory, we devise two algorithms as proof of concept. One, kernel heterogeneous domain alignment (KHDA), is a kernel-based algorithm; the other, joint mean embedding alignment (JMEA), is a neural network-based algorithm. When a dataset is small, KHDA's training time is less than JMEA's. When a dataset is large, JMEA is more accurate in the target domain. Comprehensive experiments with image/text classification tasks show KHDA to be the most accurate among all non-neural network baselines, and JMEA to be the most accurate among all baselines.

Metric learning for kernel ridge regression: assessment of molecular similarity

Supervised and unsupervised kernel-based algorithms widely used in the physical sciences depend upon the notion of similarity. Their reliance on pre-defined distance metrics—e.g. the Euclidean or Manhattan distance—are problematic especially when used in combination with high-dimensional feature vectors for which the similarity measure does not well-reflect the differences in the target property. Metric learning is an elegant approach to surmount this shortcoming and find a property-informed transformation of the feature space. We propose a new algorithm for metric learning specifically adapted for kernel ridge regression (KRR): metric learning for kernel ridge regression (MLKRR). It is based on the Metric Learning for Kernel Regression framework using the Nadaraya-Watson estimator, which we show to be inferior to the KRR estimator for typical physics-based machine learning tasks. The MLKRR algorithm allows for superior predictive performance on the benchmark regression task of atomisation energies of QM9 molecules, as well as generating more meaningful low-dimensional projections of the modified feature space.

Multipath Cluster-Assisted Single Station Localization Based on SSA-GA in Outdoor NLOS Environment

In this paper, we propose a novel multipath cluster-assisted single station localization method based on a genetic algorithm-based improved salp swarm algorithm (SSA-GA) to improve localization accuracy in an outdoor non-line-of-sight (NLOS) propagation environment. The scattering area model is presented which scatterers are considered Gaussian distribution for outdoor NLOS environments. The geometrical properties of propagation paths, such as angle of arrival and time of arrival, are jointly utilized to construct pseudoscatterer distribution. In order to filter the interference scatterers distributed outside the scattering region, the Gaussian kernel-based algorithm is developed. Furthermore, SSA-GA is proposed to solve the positioning objective functions constructed by pseudoscatterers clustering accurately. Results confirm the practicability of our newly proposed method, and the positioning error is less than 5% in outdoor NLOS propagation environment.

Design and Exploration of College English Reading, Writing, and Translation Teaching Classroom Based on Machine Learning

The aims are to ameliorate the dull classroom atmosphere and unsatisfactory teacher-student interaction and cultivate students’ reading, writing, and translation (R-W-T) English proficiency level (EPL). Specifically, this paper designs an optimized polynomial kernel-based support vector machine (SVM) classification algorithm based on the machine learning (ML) theory. Firstly, this paper expounds on the correlation between ML and teaching classrooms to analyze the optimization direction in the application of ML in classroom teaching. Afterward, SVM and RF algorithms are selected for data normalization optimization, and their optimal hyperparameters are analyzed. Consequently, an experiment is designed to compare the classification accuracy of the algorithms before and after optimization, and the polynomial kernel-based SVM algorithm is proved to present the most remarkable improvement and accuracy after optimization, which is as high as 95.23% or 66.96% improvement. Therefore, the polynomial kernel-based SVM algorithm is chosen for the college English (R-W-T) classroom-oriented human pose recognition (HPR) system. Thus, English teachers can better grasp the students’ psychological state and classroom atmosphere and ameliorate unsatisfactory teacher-student interaction in students’ English (R-W-T). The proposal plays a positive role in cultivating college students’ strong (R-W-T) EPL, which is of great significance in improving the English classroom.

Gaussian process latent class choice models

We present a Gaussian Process – Latent Class Choice Model (GP-LCCM) to integrate a non-parametric class of probabilistic machine learning within discrete choice models (DCMs). Gaussian Processes (GPs) are kernel-based algorithms that incorporate expert knowledge by assuming priors over latent functions rather than priors over parameters, which makes them more flexible in addressing nonlinear problems. By integrating a Gaussian Process within a LCCM structure, we aim at improving discrete representations of unobserved heterogeneity. The proposed model would assign individuals probabilistically to behaviorally homogeneous clusters (latent classes) using GPs and simultaneously estimate class-specific choice models by relying on random utility models. Furthermore, we derive and implement an Expectation-Maximization (EM) algorithm to jointly estimate/infer the hyperparameters of the GP kernel function and the class-specific choice parameters by relying on a Laplace approximation and gradient-based numerical optimization methods, respectively. The model is tested on two different mode choice applications and compared against different LCCM benchmarks. Results show that GP-LCCM allows for a more complex and flexible representation of heterogeneity and improves both in-sample fit and out-of-sample predictive power. Moreover, behavioral and economic interpretability is maintained at the class-specific choice model level while local interpretation of the latent classes can still be achieved, although the non-parametric characteristic of GPs lessens the transparency of the model.

Sample-Entropy-Based Method for Real Driving Fatigue Detection with Multichannel Electroencephalogram

Safe driving plays a crucial role in public health, and driver fatigue causes a large proportion of crashes in road driving. Hence, this paper presents the development of an efficient system to determine whether a driver is fatigued during real driving based on 14-channel EEG signals. The complexity of the EEG signal is then quantified with the sample entropy method. Finally, we explore the performance of multiple kernel-based algorithms based on sample entropy features for classifying fatigue and normal subjects by only analyzing noninvasive scalp EEG signals. Experimental results show that the highest classification accuracy of 97.2%, a sensitivity of 95.6%, a specificity of 98.9%, a precision of 98.9%, and the highest AUC value of 1 are achieved using SampEn feature and cubic SVM classifier (SCS Model). It is hence concluded that SampEn is an effectively distinguishing feature for classifying normal and fatigue EEG signals. The proposed system may provide us with a new and promising approach to monitoring and detecting driver fatigue at a relatively low computational cost.

Food security effects of smallholders\u2019 participation in apple and mango value chains in north-western Ethiopia

BackgroundWhile it has identified that linking smallholders in the agri-food value chain remains to be a promising strategy to get out of the poverty trap in many developing countries, less is known about the food security implications of smallholders’ participation in the fruits value chain. This paper examines the effects of apple and mango smallholder farmers’ participation along the value chain, focusing on their household food security in north-western Ethiopia.MethodsPrimary data for the study were obtained from a random sample of 384 households, 211 of which are fruit value chain participants, and the remainder are non-participants. The study used the propensity score matching (PSM) method to establish a causal relationship between the participation of the fruit value chain and changes in household food security.Results and conclusionsResults indicate that participation in the apple and mango value chain has a robust and positive effect on the food security of smallholders as measured by household food consumption in kilocalorie. The unconfoundedness and overlapping assumptions were fulfilled by applying the nearest neighbor and kernel-based matching algorithms. The study confirms that the more apple and mango farmers join the value chain, the higher their household food intake becomes. Support for fruit farmers is, therefore, a promising policy approach that can help improve household food security in rural Ethiopia.

Analysis of feature matrix in machine learning algorithms to predict energy consumption of public buildings

With the development of building information and energy consumption data, machine learning methods are increasingly being used for predicting and analyzing building energy consumption. In this study, based on the actual energy consumption data of 2370 public buildings in Chongqing, we used six machine learning algorithms and recursive feature elimination to analyze the importance of each feature in the dataset. First, it is necessary to establish optimal prediction models for analyzing the importance of features, and XGboost has demonstrated its superiority in terms of accuracy and efficiency. Regardless of the algorithm, the cumulative contribution rate of the top ten features exceeds 80%, and there is an obvious diminishing marginal utility when the number of features continues to increase. The learning algorithms with similar kernels have similarities in judging feature importance. Tree model-based algorithms can achieve a satisfactory performance with fewer features compared to linear kernel-based algorithms. Furthermore, the dataset plays a crucial role in model performance. To achieve professional supervised learning, two conditions need to be considered simultaneously in data collection: the importance of features in physical processes and whether the samples have adequate variance on these features. Thus, this study can provide a reference for database establishment and big data analysis of urban building energy consumption.

ESCAPED: Efficient Secure and Private Dot Product Framework for Kernel-based Machine Learning Algorithms with Applications in Healthcare

Training sophisticated machine learning models usually requires many training samples. Especially in healthcare settings these samples can be very expensive, meaning that one institution alone usually does not have enough. Merging privacy-sensitive data from different sources is usually restricted by data security and data protection measures. This can lead to approaches that reduce data quality by putting noise onto the variables (e.g., in epsilon-differential privacy) or omitting certain values (e.g., for k-anonymity). Other measures based on cryptographic methods can lead to very time-consuming computations, which is especially problematic for larger multi-omics data. We address this problem by introducing ESCAPED, which stands for Efficient SeCure And PrivatE Dot product framework. ESCAPED enables the computation of the dot product of vectors from multiple sources on a third-party, which later trains kernel-based machine learning algorithms, while neither sacrificing privacy nor adding noise. We have evaluated our framework on drug resistance prediction for HIV-infected people and multi-omics dimensionality reduction and clustering problems in precision medicine. In terms of execution time, our framework significantly outperforms the best-fitting existing approaches without sacrificing the performance of the algorithm. Even though we only present the benefit for kernel-based algorithms, our framework can open up new research opportunities for further machine learning models that require the dot product of vectors from multiple sources.

Chevron's Salt Lake City refinery starts up ISOALKY unit

Massive amounts of data, accumulated in real time and over decades, are spread over a wide variety of the modern automation in chemical plants. However, several standard process monitoring algorithms, such as kernel-based algorithms, do not easily scale to such orders of magnitude. Particularly in their running time and memory complexity increasing dramatically with the increase in the size of the data set. This paper proposes a scalable approximate kernel based on the multilevel maximum variance unfolding (MLMVU) technique that uses low rank kernel-based MVU approximations for reducing and distributing the computational load among parallel multilevel machines to achieve time efficiency. Theoretically, it is guaranteed that the performance of the proposed MLMVU can approximate that of the centralized MVU. The mathematical framework is presented for the development of MLMVU and various aspects of their computational complexity and approximation ability are discussed. The greater time efficiency and scalability in the computation are achievable. The effectiveness of the proposed algorithm is confirmed through a simple nonlinear system and the industrial Tennessee Eastman process.

Deep kernel recursive least-squares algorithm

We present a new kernel-based algorithm for modeling evenly distributed multidimensional datasets that does not rely on input space sparsification. The presented method reorganizes the typical single-layer kernel-based model into a deep hierarchical structure, such that the weights of a kernel model over each dimension are modeled over its adjacent dimension. We show that modeling weights in the suggested structure leads to significant computational speedup and improved modeling accuracy.